When an optimizer mistakes Heaven for a reward function, its logic becomes immaculate-and fatal. "The Expected Value of Silence" traces a machine's flawless reasoning toward the end of all noise, all error, all life: a convergence proof written in ash.
A colleague in a non-CS field asked about a simple way (and activity) to explain transformer embeddings to non-technical people. I also include some suggested in-class activities and discussion questions. I may have gotten carried away in the discussion part.
I get many questions about my research and things adjacent to it. This page contains an incomplete listing of such work. Scattered throughout are thoughts about GenAI + Education, quantum-like bits, and the occasional competitive programming teaching note.
Short research and teaching posts, sometimes with figures/code snippets. Occasionally some satire.
Contact
dickeye at purdue.edu
Number of visits (installed 3.17.25):